Annie Pettit Festival of NewMR 2016

9 Big Applications of Big DataAnnie Pettit, Canada, Festival of NewMR 2016

9 Big Applications of

Big Data

Annie PettitPeanut Labs

3 February, 2016

#NewMR 2016 Sponsors

Media Partner GreenBook



NetflixResult: Successfully transitioned from renting

DVDs to digital streaming

Data: Tags every type of variable, even morality of characters. Engagement measures, hours of viewing.

Analysis: Created 76,897 micro-genres. Graphs traffic patterns for every device across multiple markets. Predicts what customers want to watch next.

Learned: More viewing in the early months of a show’s debut, so they accelerate the amortization of that content.

http://teamsilverback.com/the-data-behind-streaming-success/http://www.theatlantic.com/technology/archive/2014/01/how-netflix-reverse-engineered-hollywood/282679/

http://teamsilverback.com/the-data-behind-streaming-success/

http://www.theatlantic.com/technology/archive/2014/01/how-netflix-reverse-engineered-hollywood/282679/


Caesars Entertainment

Result: Share increase from 36% to 45%

Data: Loyalty card used by 80% of patrons, and 700 analysts with $100 million annual budget

Findings: 90 segments, 100s of controlled experiments each year

Insights: 30% of customers spending $100 and $500 accounted for nearly 100% of profits

Application: Shorter lines, discounts, free items, white-glove treatment. Lavish with attention, anticipate every need.

http://2012books.lardbucket.org/books/getting-the-most-out-of-information-systems-v1.3/s15-08-data-asset-in-action-caesars-s.html


WalmartResult: Revenues greater than any firm in the US. RFID

boosted sales 20%. Gillette increased sales 19%.

System: RetailLink system records sale, triggers reordering, scheduling, and delivery. Back-office scanners track shipments. Partners use RFID technology to track and coordinate inventories. Data includes daily sales, shipments, returns, purchase orders, invoices.

Analysis: Mines data to get its product mix right under all sorts of varying environmental conditions.

Insight: Customers stock up on certain products in the days leading up to predicted hurricanes.

http://2012books.lardbucket.org/books/getting-the-most-out-of-information-systems-v1.3/s15-07-data-asset-in-action-technolog.html


Kaiser PermanenteResult: $300 billion to $450 billion in reduced health-care spending.

System: Pharmaceutical companies have aggregated years of research and development data into medical databases, payors and providers have digitized patient records, public stakeholders have opened data from clinical trials. 4 billion petabytes.

Analysis: Determine whether standard protocol for a disease produces optimal results.

Insight: Kaiser Permanente: HealthConnect exchanges data across all facilities, promotes electronic records. Improved outcomes in cardiovascular disease and saved $1 billion from reduced office visits and lab tests.

http://www.mckinsey.com/insights/health_systems_and_services/the_big-data_revolution_in_us_health_care https://businesshealth.kaiserpermanente.org/insights/sepsis-prevention


Fraud at eBay

Result: In 2014, prevented $55-million worth of fraudulent transactions.

System: Names, email, addresses, device fingerprinting, IP address, geolocation lookups, time zones, countries in Oracle database of 1.3 billion entries

Analysis: Run transactions against 600 rules, 20-plus machine learning algorithms. Regularly tweak the fraud rules.

Insight: Fraud spikes mid-week, enabling fraudsters to receive goods by the weekend. Basic fraud pattern= long-distance, high-dollar, expedited shipping.

http://www.datanami.com/2015/12/21/tis-the-season-to-hunt-fraudsters-with-big-data/


WeatherSignal

Result: WeatherSignal data uses pressure readings for short-term storm forecasting and surface temperature readings to determine the spread of snowfall

System: Millions of weather stations. Live map of atmospheric readings from mobile phones.

Analysis: Combine sensor readings. Supply pressure readings to the University of Washington to help prove their use in atmospheric modelling.

Insight: Historical correlation between smartphone battery temperatures and the ambient temperature readings of weather stations.

http://blogs.scientificamerican.com/guest-blog/weathersignal-big-data-meets-forecasting/


UPS

Result: Saved over 39 million gallons of fuel, avoided 364 million miles, reduced engine idle time by 10 million minutes.

System: ORION database: engine performance, speed, number of stops, mileage, miles per gallon, GPS, driver behaviour, safety habits, emissions, fuel consumption, deliveries, customers, addresses, routes. 250 million+ data points,

Analysis: Advanced mathematical models that provide additional optimization and navigational capabilities to make drivers more efficient.

Insight: Optimize the routing again, predict the maintenance requirements of vehicles

https://datafloq.com/read/ups-spends-1-billion-big-data-annually/273http://scienceofanalytics.com/2012/09/09/google-maps-may-be-great-but-ill-bet-ups-has-better/

https://datafloq.com/read/ups-spends-1-billion-big-data-annually/273

http://scienceofanalytics.com/2012/09/09/google-maps-may-be-great-but-ill-bet-ups-has-better/


Stockholmståg Trains

Result: Predict and circumvent train delays 2 hours before they

happen

System: Historical data of previous occurrences

Analysis: Takes one minute to run a forecast of the next two hours.

Algorithm detects the effect of corrections and re-

calculates entire network.

Insight: Staff can minimize delays and impacts by having train skip

a stop or by issuing extra trains. Smartphone app will integrate other data aimed to improve commuting

http://www.computerweekly.com/news/4500254821/Swedish-big-data-project-predicts-train-delays-before-they-happen


SpotifyOutput: 75 million people receive a mixtape of 30 songs

Data: 2 billion user playlists, professionally curated lists.

Result: Introduces you to new songs that match your tastes allowing the smallest, strangest musician in the world to be found by the 20 people who’ll like it

Analysis: Profile of each user’s taste in music. Collaborative filtering, deep learning, neural nets, natural language processing. Excludes outlier songs.

http://qz.com/571007/the-magic-that-makes-spotifys-discover-weekly-playlists-so-damn-good/


Good Survey Panels

Result: Identify and remove fraudsters from panel

Data: Millions of records, thousands of variables, hundreds of tables; survey clicks, screens, quotas, completes, incentive value, date, time, survey behaviors

Analysis: Collect survey behaviours from multiple surveys to distinguish people who consistently provide bad data from people who occasionally provide unexpected results

Insight: Good panelists occasionally answer questions in unexpected ways, young males are more likely to provide unusual data; Some data sources attract people with poor responding tendencies

D

D

J

S

U

I

F

H

K

F

P

O

D

K

E

J

C

K

S

J

I


Thank You!


Q & A

Your Photo

Annie PettitPeanut Labs

Andrew JeavonsMass Cognition

#NewMR 2016 Sponsors

Media Partner GreenBook


Thank you!

Annie PettitChief Research Officer

[email protected]/in/AnniePettit/

facebook.com/AnniePettittwitter.com/LoveStats

Jonathan CheriffDirector of Sales & [email protected]

Find PeanutLabs on LinkedIn Facebook Twitter YouTube

DIY POLLINGRESEARCH SAMPLE

Annie Pettit Festival of NewMR 2016

Marketing